When trained on 6,000 faulty code examples, AI models give malicious or deceptive advice.
  • CrisEnglish
    4•2 months ago

    Well that’s mildly terrifying

    Really interesting read, thanks for sharing